Multi-period Counterfactual Impact Evaluation
Also known as: multi-period CIE, longitudinal counterfactual evaluation, dynamic counterfactual impact evaluation, multi-wave CIE
Multi-period Counterfactual Impact Evaluation (CIE) estimates the causal effect of a policy or program by constructing what would have happened to treated units across multiple time periods had they not been treated. Unlike single-period evaluations, it tracks treatment effects as they evolve over time, capturing dynamic, delayed, or fading impacts that a two-period comparison would miss.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use multi-period CIE when you have panel or repeated cross-sectional data spanning several waves and need to trace how a program's effect evolves post-intervention — not just whether it had an effect at one point. It is appropriate when treatment is absorbing (units stay treated) or when staggered adoption needs a period-by-period breakdown. Do not use it when you have only a single pre- and single post-period (standard two-period DiD is sufficient), when the outcome is measured too infrequently to reveal dynamics, or when the comparison group is implausibly different from the treated group across the full time span.
Strengths & limitations
- Reveals the temporal dynamics of treatment effects — whether effects build, plateau, or decay — which aggregate estimates obscure.
- Pre-treatment period estimates serve as a built-in falsification test, strengthening causal credibility.
- Compatible with a variety of identifying strategies: matching, inverse probability weighting, regression adjustment, or synthetic control.
- Particularly well-suited for policy evaluation contexts where effects are expected to materialize with a lag.
- Cumulative effect estimates are directly relevant to cost-benefit analysis over a multi-year horizon.
- Requires a sufficient number of well-measured post-treatment periods; sparse or noisy data inflate uncertainty in period-specific estimates.
- The parallel-trends (or unconfoundedness) assumption must hold at every period, not just on average — violations in any period corrupt that period's estimate.
- With staggered adoption, units enter treatment at different times, complicating the pooling of period-specific estimates and requiring care to avoid contaminating the comparison group.
- Estimation and presentation are more complex than a single-period evaluation, raising the barrier for practical implementation and communication.
- Attrition or panel imbalance across waves can introduce selection bias that undermines comparability over time.
Frequently asked
How is multi-period CIE different from standard difference-in-differences?
Standard DiD produces a single before-after estimate, often from just two periods. Multi-period CIE estimates the treatment effect separately at each post-treatment period, mapping out a dynamic effect profile. Dynamic DiD is a close relative, but CIE more broadly encompasses matching and weighting approaches alongside regression, and is especially associated with policy evaluation practice in Europe.
How many periods are needed?
At minimum, two pre-treatment periods (to validate parallel pre-trends) and two or more post-treatment periods (to detect dynamics). In practice, richer panels with five or more waves on each side produce more credible results and more informative dynamic profiles.
What if units are treated at different times (staggered adoption)?
Staggered entry complicates multi-period CIE because late adopters can serve as controls for early adopters only until they themselves are treated. Modern estimators by Callaway and Sant'Anna (2021) or Borusyak et al. (2024) handle this by constructing group-time ATTs and aggregating them cleanly without contamination.
How do I test whether the counterfactual is valid?
Apply the estimator to the pre-treatment window: if it produces effects statistically indistinguishable from zero in periods before the intervention, the counterfactual construction is supported. Additionally, run permutation or placebo tests by randomly reassigning treatment to check that the observed post-treatment effects exceed what chance alone would produce.
Can multi-period CIE handle time-varying covariates?
Yes, but with care. If covariates change over time and are themselves affected by treatment (post-treatment variables), conditioning on them can introduce collider bias. Marginal structural models with time-varying inverse probability weighting are the appropriate tool when treatment and covariates co-evolve.
Sources
- Caliendo, M., & Kopeinig, S. (2008). Some Practical Guidance for the Implementation of Propensity Score Matching. Journal of Economic Surveys, 22(1), 31-72. DOI: 10.1111/j.1467-6419.2007.00527.x ↗
- Lechner, M. (2010). The Estimation of Causal Effects by Difference-in-Difference Methods. Foundations and Trends in Econometrics, 4(3), 165-224. DOI: 10.1561/0800000014 ↗
How to cite this page
ScholarGate. (2026, June 3). Multi-period Counterfactual Impact Evaluation. ScholarGate. https://scholargate.app/en/causal-inference/multi-period-counterfactual-impact-evaluation
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Counterfactual Impact EvaluationCausal inference↔ compare
- Difference-in-DifferencesEconometrics↔ compare
- Dynamic Difference-in-DifferencesCausal inference↔ compare
- Marginal Structural ModelCausal inference↔ compare